Education and work experience
July 2010 - present : Research engineer at Twenga, Paris, France
2006 – 2009 : PhD student in Computer Science, University of Rennes 1, France
Research area: Automatic audiovisual stream structuring
Keywords: TV broadcast, video indexing, clustering, machine learning, video classification, inductive logic programming
Research conducted at and funded by Orange Labs – France Télécom R&D,
Cesson Sévigné, France, in collaboration with TEXMEX team (IRISA/INRIA),
Rennes, France
Supervisors: Sid-Ahmed Berrani (Orange Labs), Patrick Gros (IRISA)
September 2008 : Summer School on Multimedia Semantics, one week, Chania, Crete, Greece
2005 – 2006 : Master of Science in Computer Science, with a major in Image and Computer Vision, National Institute for Applied Sciences (INSA) of Lyon, France
2003 – 2006 : Master of Engineering in Computer Science, National Institute for Applied Sciences (INSA) of Lyon, France
PhD Thesis : Automatic structuring of television streams
Defended at IRISA, Rennes, on Jully 13th, 2010
Abstract
TV streams are structured : they consist of consecutive television programs (news , movies, magazines, etc.) and of inter-programs (commercial, trailer, sponsoring, etc.). As soon as they are broadcasted on the air, the TV streams lose unfortunately all information on their structure. The main problem of this thesis is to automatically recover, from the received linear audio/visual signals and using the possible available metadata provided by TV channels, the original structure of the TV streams, with the exact start and the exact end of each broadcasted program.
This thesis presents a full system for automatic TV stream structuring that has been rigorously evaluated on four weeks of real TV streams, on two different TV channels. The system makes use of the repetition property of the inter-programs. A first segmentation step based on repeated video sequence detection is performed. Resulting segments (the occurrences of repeated sequences and the rest of the stream) are then classified in program segments or in inter-program segments. The proposed classification solution is based on Inductive Logic Programming. It makes use of neighboring relations between segments. Finally, detected program segments are labelled using metadata and those that belong to the same TV program are reunified. When metadata are not available, program segments of the same TV program can only be reunified based on visual and structure similarities.
Keywords
Wire broadcasting, Television programs, Video recordings, Automatic indexing, Audiovisual repetitions, Machine learning, Logic programming
Thèse : Délinéarisation automatique de flux de télévision
Soutenue à l'IRISA, Rennes, le 13 juillet 2010 : manuscrit [pdf]
, présentation [pdf]
Résumé
Les flux de télévision sont structurés : ils sont en effet composés de programmes successifs (journaux, films, magazines, etc.) et entrecoupés par des inter-programmes (publicités, bandes annonces, parrainages, etc.). Dès que les flux sont diffusés sur les ondes, ils perdent malheureusement toute information de structure. La problématique de la délinéarisation automatique est de retrouver la structure des flux TV, avec en particulier le début précis et la fin précise de chaque programme, à partir des signaux audiovisuels reçus et des métadonnées éventuellement fournies par les chaînes TV.
Cette thèse présente un système complet de délinéarisation automatique rigoureusement évalué sur quatre semaines de flux TV réels, pour deux chaînes de télévision différentes. Les travaux se basent sur la propriété de répétition des inter-programmes. Cette propriété est exploitée à travers la détection de toutes les répétitions d'un flux grâce à une technique de clustering des images clés du flux. Ces répétitions servent à la création de segments qui sont ensuite classés en segments de programme ou en segments d'inter-programme suivant les caractéristiques des répétitions et les relations entre les segments. Pour cela, le système utilise la programmation logique inductive. Une fois les segments classés, les segments de programme appartenant à un même programme sont étiquetés et réunifiés grâce aux métadonnées éventuelles. En l'absence de métadonnées, les segments de programme d'un même programme peuvent être seulement réunifiés grâce à des similarités visuelles.
Mots clés
Télédiffusion, Télévision - Émissions, Videos, Indexation automatique, Répétitions audiovisuelles, Apprentissage automatique, Programmation logique
Publications and Patents
Journal
- G. Manson and S.-A. Berrani. “Automatic TV Broadcast Structuring”. International Journal of Digital Multimedia Broadcasting, special issue on “Video Analysis, Abstraction and Retrieval: Techniques and Applications”, vol. 2010, Article ID 153160, 16 pages, 2010
- S.-A. Berrani, G. Manson, and P. Lechat. “A non supervised approach for repeated sequence detection in TV broadcast streams”. Signal Processing: Image Communication, special issue on “Semantic Analysis for Interactive Multimedia Services”, 23(7):525–537, 2008[pdf] *
Patents
- G. Manson and S.-A. Berrani. “Process for TV program segment fusion", filed by France Télécom on February 6th, 2009
- S.-A. Berrani and G. Manson. “Process for automatic classification of TV broadcast stream segments", filed by France Télécom on February 15th, 2008
International Conferences
- G. Manson and S.-A. Berrani. “Content-Based Video Segment Reunification for TV Program Extraction”. In Proc. of the IEEE Int. Symp. on Multimedia, San Diego, California, USA, December 2009
- G. Manson and S.-A. Berrani. “Repetition Density-Based Approach For TV Program Extraction”. In Proc. of the IEEE Int. Work. on Image Analysis for Multimedia Interactive Services, London, UK, May 2009[pdf] *
- G. Manson and S.-A. Berrani. “TV Broadcast Macro-Segmentation using the Repetition Property of Inter-Programs”. In Proc. of the IASTED Int. Conf. on Signal Processing: Pattern Recognition and Applications, Innsbruck, Austria, February 2009[pdf] *
- G. Manson and S.-A. Berrani. “An Inductive Logic Programming-Based Approach for TV Stream Segment Classification”. In Proc. of the IEEE Int. Symp. on Multimedia, Berkeley, California, USA, December 2008[pdf] *
- S.-A. Berrani, P. Lechat, and G. Manson. “TV broadcast macro-segmentation: Metadata-based vs content-based approaches”. In Proc. of the ACM Int. Conf. on Image and Video Retrieval, Amsterdam, The Netherlands, July 2007[pdf] *
French Conference
- G. Manson and S.-A. Berrani. “Structuration automatique de flux télévisés”. In Proc. of the GRETSI Symposium on Signal and Image Processing, Dijon, France, September 2009
Demonstration Papers
- G. Manson, X. Naturel and S.-A. Berrani. “Automatic Program Extraction From TV Streams”. In Proc. of the ACM European Interactive TV Conf., Leuven, Belgium, June 2009
- G. Manson, X. Naturel and S.-A. Berrani. “Online Macro-segmentation of Television Streams”. In Proc. of the ACM Int. Multimedia Modeling Conf., Sophia-Antipolis, France, January 2009
- G. Manson and S.-A. Berrani. “A TV Stream Macro-segmentation System”. In Proc. of the IEEE Int. Symp. on Multimedia, Berkeley, California, USA, December 2008[pdf] *
Teaching activities
Introduction to C++ programming language
26 hours : 13 hours of lab classes held in 2008 and 13 hours held in 2009 at the National School for Statistics and Information Analysis (ENSAI), Rennes, France
Graphical User Interface programming with wxWidgets
16 hours : 4 hours of lecture classes and 12 hours of lab classes held in 2009 at the Institut de Formation Supérieure en Informatique et Communication (IFSIC) in University of Rennes 1, Rennes, France
* Copyright notice
This material is the author's version of the work. It is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.
